Korsgaard famously distinguished two senses of 'intrinsic value': (1) final (non-instrumental) value, and (2) value held in virtue of intrinsic properties. Mementos or other objects with symbolic significance, for example, may be valued non-instrumentally precisely for their relational properties: they connect us to someone or something we care about, and we value this connection in its own right, not as a means to any other end. So far so good. But I also want to distinguish the question of an object's final value from a further question:
(3) Whether the object is of ultimate concern (a source of reasons, on whose behalf the reason exerts its normative force), or whether it instead contributes value to some 'larger' entity that is our real concern (that for the sake of which we seek value).
Even if a memento has final value, it surely isn't of ultimate concern -- it is not an entity whose interests we serve. So mementos might be said to lack the 'intrinsic value' of persons in the sense that they are valuable for our sakes, not their own. (World-consequentialists would go even further and say that the reason to make persons better off is for the sake of making the world better. It's the world as a whole, not its particular constituents, that is of ultimate concern or 'intrinsic value' in this third sense.)
P.S. It's an interesting question whether final value is the same thing as contributory value. Perhaps we need to add a further question yet:
#(4) Whether, holding all else strictly equal, the addition of the object makes the world a better place.
Suppose average utilitarianism is true. Then a life might have positive welfare value but negative contributory value (because below average). Since the life has positive welfare value, we can say that it is non-instrumentally good in respect of its intrinsic features. Does that mean the life has final value but not contributory value? Well, not necessarily. After all, the life also has extrinsic features in virtue of which it is (ex hypothesi) non-instrumentally bad. So it looks like this fourth question is redundant in light of the first, i.e. whether the object is non-instrumentally desirable. We simply need to take care to distinguish what's desirable all things considered from more abstracted questions, e.g. what's desirable in some respect, or for Bob's sake, etc.
Update: further discussion of this 4th possibility here.